Update

edit_app.py

@@ -53,42 +53,28 @@ pipe = StableDiffusionInstructPix2PixPipeline.from_pretrained(
 example_image = Image.open("imgs/example.jpg").convert("RGB")
 
 
-def load_example(
-    steps: int,
+def randomize(
     randomize_seed: bool,
     seed: int,
     randomize_cfg: bool,
     text_cfg_scale: float,
     image_cfg_scale: float,
-):
-    example_instruction = random.choice(example_instructions)
-    return [example_image, example_instruction] + generate(
-        example_image,
-        example_instruction,
-        steps,
-        randomize_seed,
-        seed,
-        randomize_cfg,
-        text_cfg_scale,
-        image_cfg_scale,
-    )
+) -> tuple[int, float, float]:
+    seed = random.randint(0, 100000) if randomize_seed else seed
+    text_cfg_scale = round(random.uniform(6.0, 9.0), ndigits=2) if randomize_cfg else text_cfg_scale
+    image_cfg_scale = round(random.uniform(1.2, 1.8), ndigits=2) if randomize_cfg else image_cfg_scale
+    return seed, text_cfg_scale, image_cfg_scale
 
 
 def generate(
     input_image: Image.Image,
     instruction: str,
     steps: int,
-    randomize_seed: bool,
     seed: int,
-    randomize_cfg: bool,
     text_cfg_scale: float,
     image_cfg_scale: float,
     progress=gr.Progress(track_tqdm=True),
-):
-    seed = random.randint(0, 100000) if randomize_seed else seed
-    text_cfg_scale = round(random.uniform(6.0, 9.0), ndigits=2) if randomize_cfg else text_cfg_scale
-    image_cfg_scale = round(random.uniform(1.2, 1.8), ndigits=2) if randomize_cfg else image_cfg_scale
-
+) -> Image.Image:
     width, height = input_image.size
     factor = 512 / max(width, height)
     factor = math.ceil(min(width, height) * factor / 64) * 64 / min(width, height)
@@ -108,7 +94,37 @@ def generate(
         num_inference_steps=steps,
         generator=generator,
     ).images[0]
-    return [seed, text_cfg_scale, image_cfg_scale, edited_image]
+    return edited_image
+
+
+def load_example(
+    steps: int,
+    randomize_seed: bool,
+    seed: int,
+    randomize_cfg: bool,
+    text_cfg_scale: float,
+    image_cfg_scale: float,
+    progress=gr.Progress(track_tqdm=True),
+):
+    example_instruction = random.choice(example_instructions)
+    seed, text_cfg_scale, image_cfg_scale = randomize(
+        randomize_seed, seed, randomize_cfg, text_cfg_scale, image_cfg_scale
+    )
+    return [
+        example_image,
+        example_instruction,
+        seed,
+        text_cfg_scale,
+        image_cfg_scale,
+        generate(
+            example_image,
+            example_instruction,
+            steps,
+            seed,
+            text_cfg_scale,
+            image_cfg_scale,
+        ),
+    ]
 
 
 def reset():
@@ -116,7 +132,7 @@ def reset():
 
 
 def process_example(input_image: Image.Image, instruction: str, seed: int) -> Image.Image:
-    return generate(input_image, instruction, 50, False, seed, False, 7.5, 1.5)[-1]
+    return generate(input_image, instruction, 50, seed, 7.5, 1.5)
 
 
 with gr.Blocks() as demo:
@@ -214,18 +230,28 @@ InstructPix2Pix: Learning to Follow Image Editing Instructions
             text_cfg_scale.submit,
             image_cfg_scale.submit,
         ],
+        fn=randomize,
+        inputs=[
+            randomize_seed,
+            seed,
+            randomize_cfg,
+            text_cfg_scale,
+            image_cfg_scale,
+        ],
+        outputs=[seed, text_cfg_scale, image_cfg_scale],
+        queue=False,
+        api_name=False,
+    ).then(
         fn=generate,
         inputs=[
             input_image,
             instruction,
             steps,
-            randomize_seed,
             seed,
-            randomize_cfg,
             text_cfg_scale,
             image_cfg_scale,
         ],
-        outputs=[seed, text_cfg_scale, image_cfg_scale, edited_image],
+        outputs=edited_image,
         api_name="run",
     )